A Novel Hybrid Models Based on Gaussian Process Regression for Short-Term Wind Power Forecasting

Short-term wind power prediction with high accuracy and robustness is important for stable operation and effective dispatch of wind power systems. Nevertheless, it is relatively difficult to obtain satisfactory prediction results in wind power systems due to the nonlinearity and instability of wind power generation. In this paper, we propose a high accuracy combined wind power prediction model using a least squares support vector machine (LSSVM) optimized by an improved whale optimization algorithm (IWOA) and a long-short time memory (LSTM) improved by a Bayesian optimization algorithm (BO) as two base models to predict wind power time series, respectively. A Gaussian process regression (GPR) model is used to combine the independent forecasting results generated by the two forecasting models. Comparing the proposed prediction model with the two base models based on the National Renewable Energy Laboratory (NREL) public wind field dataset, the combined model significantly improves the prediction precision.